Development of a Self-Adaptive Ant Colony Optimization for Designing Pipe Networks

Optimum design of a water distribution network leads to a nonlinearly constrained problem that is inherently discrete and multimodal. The penalty function approach is traditionally used to handle the problem constraint. Constraint violations are directly added to the objective function by applying the penalty functions. Penalty functions and coefficients are case-dependent requiring careful attention to be appropriately justified and calibrated. This is a complicated issue that requires many initial trial-and-error computations and may deform the problem search space. To overcome the problems associated with the penalty function method, this study introduces a self-adaptive ant colony optimization (SACO) method. The focus of this study is on developing a methodology for adaptive handling of the problem constraints without using penalty functions. The proposed approach is applied to a benchmark example, i.e., Hanoi pipe network. According to a comparing between the results of the new method and the conventional penalty function method, the self-adaptive scheme would remarkably increase the optimization efficiency as well as chances of reaching a global optimum design. Besides, there is no need to determine and calibrate any coefficient for meeting the problem constraints.